uzmorph-bigru 0.1.1

Uzbek Neural Morphological Analyzer (BiGRU Architecture)

uzmorph-bigru: Uzbek Neural Morphological Analyzer (BiGRU Architecture)

uzmorph-bigru is a high-accuracy word-level morphological analyzer for the Uzbek language. It leverages the Bidirectional Gated Recurrent Unit (BiGRU) architecture to offer a lighter yet powerful alternative to standard LSTMs.

Performance & Use Case

• Architecture: Bidirectional GRU (2 Layers, 128 hidden units).
• Efficiency: Approximately 20% faster inference and 15% smaller model size (~0.9MB) compared to BiLSTM.
• Accuracy: >96% on common word sets.
• Ideal For: Mobile applications or environments with limited memory where high throughput is required.

Installation

pip install uzmorph-bigru

Quick Start (Usage Examples)

1. Simple Analysis (String Output)

from uzmorph_bigru.uzmorph_bigru import uzmorph_bigru

analyzer = uzmorph_bigru()
result = analyzer.analyze("kitobim")
print(result)

# Output:
# Result: 'kitobim' -> Stem: kitob | POS: NOUN | Tags: [possession=1, singular=1]

2. Structured Data (Dict/JSON)

# To Dictionary
data = analyzer.analyze("maktabda").to_dict()
# { "word": "maktabda", "stem": "maktab", "pos": "NOUN", "cases": "Locative" }

# To JSON
json_string = analyzer.analyze("yozyapmiz").to_json()

Supported Tags & Features

Part of Speech (POS)

• NOUN (Ot), VERB (Fe'l), ADJ (Sifat), ADV (Ravish), NUM (Son), PRN (Olmosh).

Grammatical Features

• Cases: Nominative, Ablative, Accusative, Dative, Genitive, Locative.
• Possession: 1 (Mening), 2 (Sening), 3 (Uning).
• Number: singular=1, plural=1.
• Verb Specific:
  • Tense: Past (O'tgan), Present (Hozirgi), Future (Kelasi).
  • Voice: Causative, Passive, Reciprocal, Reflexive.
  • Mood/Impulsion: Imperative, Conditional, Proposal.

License

MIT